Commit: 9b622e2bbcf0 ("raid10: increment write counter after bio is split") in
4.8 fixed a regression introduced by commit: 20d0189b1012 ("block: Introduce
new bio_split()") in 3.14. The bug could cause data corruption or hang if
raid10 resync runs, so it's suitable for stable. For some reason it's slipped
asway from marking as stable.
Thanks,
Shaohua